Free Disk Space on a Linux Server

Logs, on-disk backups and package cache fill a Linux server disk, in that order — databases grow too, but you measure those and shrink them with a planned dump, never with rm. Start one level down from / and stay on that filesystem until the heavy directory names itself.

• /var/log and the systemd journal — web, mail and database logs plus the journal grow on every busy day.
• Backups on the same disk — a plugin keeping every copy under the site, or a dump left in /root, spends the space the site needs.
• Package cache — apt, yum and dnf keep downloaded packages under /var/cache after the software is already installed.
• Databases — du -sh /var/lib/mysql or /var/lib/postgresql sizes the data directory. Never delete ibdata, .ibd files or the PostgreSQL data directory.

A size query lists schemas without freeing anything:

SELECT table_schema,
       ROUND(SUM(data_length + index_length)/1024/1024, 1) AS mb
FROM information_schema.tables
GROUP BY table_schema ORDER BY mb DESC;

Keep one copy of any backup off this server. Server backup is that copy; a file that exists only on this disk is not a backup you can safely delete.

logrotate renames, compresses and eventually deletes log files on a schedule you set, so /var/log reaches a fixed ceiling instead of growing forever — and it does this without breaking the daemon still writing to the log. A safe web-log block rotates weekly, keeps four compressed copies and skips empty or missing files.

Dry-run any new file first: logrotate -d /etc/logrotate.conf prints the plan and changes nothing. Never rm a log a daemon still holds open — logrotate can copy-and-truncate or signal the daemon to reopen the file, whereas deleting the live file leaves the daemon writing to a removed inode and df stays full until the service restarts.

Clean when Use% on / passes 80, before MySQL or mail stops — a Linux server disk is a fixed size, and "unlimited disk" is a shared-hosting slogan, not a property of this mount. A weekly SSH check of df -h / and df -i / is enough to catch the trend.

• logrotate in place for web and app logs, with a keep-count that matches a weekly cleanup.
• Backups off the box: one copy on object or backup storage, not under the site root.
• Monitor the right metric: bandwidth monitoring measures transfer, not stored bytes — it will never warn you that / is full.
• Plan the workload: if Avail stays near zero after the safe deletes, the site needs a larger disk on its VPS plan (from ₹499 a month before 18% GST, configuration quoted).

If a full disk has locked you out of SSH entirely, email support@webji.services — staffed 24/7 in Hindi and English, average first reply under 60 minutes. The office phone is not a support line. For the shared-hosting contrast, where the host manages the disk for you, see shared hosting; the SSH key guide covers the key step before any of this.